Detection of Pilot Contamination Attack for Multi-Antenna Based Secrecy Systems

This paper considers the problem of detecting the pilot contamination attack for multi-antenna secrecy systems. The detection problem is formulated as a binary hypothesis problem and the likelihood ratio or the generalized likelihood ratio is employed as a decision statistic for the detection of the pilot contamination. Using the Neyman-Pearson criterion, we then develop several detection methods under the different assumptions of the channel and noise statistics. To be specific, the cases of interest are as follows: (i) exact knowledge of all covariance matrices, (ii) exact knowledge of the legitimate channel and noise covariance matrices, but no knowledge of the jamming channel covariance matrix and (iii) no knowledge of all covariance matrices. The performance comparison of the proposed schemes with the conventional scheme is performed by simulations.